Project description:

This is an R&D position and it involves development of a solution for generating secure code (code without security vulnerabilities or bugs) and evaluating existing or generated code for presence of security issues or bad practices using LLM models.

We are using such technologies as python, torch, transformers, llama.cpp, awq, ANTLR, AWS, Sagemaker, Kubernetes, and we are open to adopting new ones if they fit the task.

Tasks will include models training and validation, setting up evaluation crit

Responsibilities:

Tasks will include models training and validation, setting up evaluation criteria, domain-specific datasets preparation and augmentation;Training, validating, building inference pipeline;Python, torch, transformers, PEFT, SFT, DPO, LLM training engines, etc;Optimizing model performance;LLM inference engines, vllm, llama.cpp, awq, etc;Automatic Prompt engineering;Parsing, preparing dataset for training – antlr4, ast-grep, code syntax tree parsing, etc;Preparing demos, visualizing results – flask, matplotlib, seaborn, etc.

Requirements:

Experience in AI of at least 4 years and overall programming experience of 8+ years;Python, torch, transformers (related keywords: pytorch, tensorflow, keras, onnx);Llama, GPT, MPT, T5, Gemma, LLM, training and inference engines;Analytical and problem-solving skills, living interest in modern model training techniques (PEFT, SFT, DPO, etc);Experience with data processing, data transformations, algorithms and data structures, rapid prototyping.

Additional skills that are good to have:

Experience deploying AI models into production environments (AWS, GCP, Azure);Experience creating software solutions from scratch to production;Experience with multiple programming languages (Java/Kotlin/Scala, C#, JavaScript/Typescript, PHP, C/C++, etc.);Secure coding expertise, understanding security vulnerabilities (CWE, OWASP, NIST);Basic knowledge of web development (backend, frontend);Performance Optimization, concurrency and parallelism, debugging and profiling tools;Signal processing, audio processing, image/video processing, computer vision.

We offer:

20 working days of paid vacation per year;Official holidays of Ukraine – days off;Modern equipment for work;Corporate events;External and internal training: conferences, professional events, courses, corporate e-learning platforms;English speaking club;Work in a respectful and supportive team.

Hiring process:

Review CV;HR interview;Professional interview.

We are looking forward to your application!